live
MODEL IDideogram:4.6@0

Ideogram 4.6

Ideogram
by
Available with Zero Data Retention

Ideogram 4.6 is Ideogram's image generation and editing model, with native output up to 4096x4096. It generates images from text at any size on a 32-pixel grid, with an optional transparent background, and edits a source image using up to four more images as references. It keeps characters recognizable across scenes and matches exact colors more reliably than Ideogram 4.5, which suits product shots, brand recoloring, and campaign variations.

Ideogram 4.6

Character consistency

How to keep one person or mascot recognizable across a series with Ideogram 4.6: new scenes from one portrait, outfit changes, two characters together and an illustrated mascot.

Introduction

Describe the same person in two prompts and you get two different people. Every generation builds the face from the words alone, so the face and the glasses drift as soon as the scene changes, and a campaign built around one spokesperson stops looking like one person.

Ideogram 4.6 takes the character from an image instead of a description. Pass a portrait as the first entry of inputs.referenceImages, describe a new scene in positivePrompt, and the model keeps the person recognizable in it.

Built from 1 reference
  • Image 1

This guide covers the request, carrying one person through a campaign, changing what she wears, putting two characters in one scene, and keeping an illustrated mascot on model.

The request

The portrait goes in as image 1, and the prompt describes a new frame around the person in it:

Try in Playground
import { createClient } from '@runware/sdk'

const client = await createClient({ apiKey: process.env.RUNWARE_API_KEY })
await client.connect()

const [result] = await client.run({
  model: 'ideogram:4.6@0',
  positivePrompt: 'Advertising photograph for a home improvement store: the woman from image 1 rolls warm terracotta paint onto a wall in a sunlit living room under renovation, half the wall already painted, a drop cloth on the floor and a stepladder beside her, wearing paint-flecked denim overalls over a white t-shirt. She looks over her shoulder at the camera, smiling. Bright morning light from a window on the left, shot on a 35mm lens at chest height. Keep her face, her hair, her glasses and her earring identical to image 1.',
  inputs: {
    referenceImages: [
      'https://assets.runware.ai/guides/ideogram-4-6/character-consistency/reference-rosa.jpg'
    ]
  },
  width: 3072,
  height: 2048
})
import asyncio
import os

from runware import Runware


async def main():
    async with Runware(api_key=os.environ["RUNWARE_API_KEY"]) as client:
        results = await client.run({
            "model": "ideogram:4.6@0",
            "positivePrompt": "Advertising photograph for a home improvement store: the woman from image 1 rolls warm terracotta paint onto a wall in a sunlit living room under renovation, half the wall already painted, a drop cloth on the floor and a stepladder beside her, wearing paint-flecked denim overalls over a white t-shirt. She looks over her shoulder at the camera, smiling. Bright morning light from a window on the left, shot on a 35mm lens at chest height. Keep her face, her hair, her glasses and her earring identical to image 1.",
            "inputs": {
                "referenceImages": [
                    "https://assets.runware.ai/guides/ideogram-4-6/character-consistency/reference-rosa.jpg"
                ]
            },
            "width": 3072,
            "height": 2048
        })


asyncio.run(main())
curl https://api.runware.ai/v1 \
  -H "Authorization: Bearer $RUNWARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '[
    {
      "taskType": "imageInference",
      "taskUUID": "9a4f1c68-2e57-4b3d-8d90-6c1e7f3a2b45",
      "model": "ideogram:4.6@0",
      "positivePrompt": "Advertising photograph for a home improvement store: the woman from image 1 rolls warm terracotta paint onto a wall in a sunlit living room under renovation, half the wall already painted, a drop cloth on the floor and a stepladder beside her, wearing paint-flecked denim overalls over a white t-shirt. She looks over her shoulder at the camera, smiling. Bright morning light from a window on the left, shot on a 35mm lens at chest height. Keep her face, her hair, her glasses and her earring identical to image 1.",
      "inputs": {
        "referenceImages": [
          "https://assets.runware.ai/guides/ideogram-4-6/character-consistency/reference-rosa.jpg"
        ]
      },
      "width": 3072,
      "height": 2048
    }
  ]'
runware run ideogram:4.6@0 \
  positivePrompt="Advertising photograph for a home improvement store: the woman from image 1 rolls warm terracotta paint onto a wall in a sunlit living room under renovation, half the wall already painted, a drop cloth on the floor and a stepladder beside her, wearing paint-flecked denim overalls over a white t-shirt. She looks over her shoulder at the camera, smiling. Bright morning light from a window on the left, shot on a 35mm lens at chest height. Keep her face, her hair, her glasses and her earring identical to image 1." \
  inputs.referenceImages.0=https://assets.runware.ai/guides/ideogram-4-6/character-consistency/reference-rosa.jpg \
  width=3072 \
  height=2048
{
  "taskType": "imageInference",
  "taskUUID": "9a4f1c68-2e57-4b3d-8d90-6c1e7f3a2b45",
  "model": "ideogram:4.6@0",
  "positivePrompt": "Advertising photograph for a home improvement store: the woman from image 1 rolls warm terracotta paint onto a wall in a sunlit living room under renovation, half the wall already painted, a drop cloth on the floor and a stepladder beside her, wearing paint-flecked denim overalls over a white t-shirt. She looks over her shoulder at the camera, smiling. Bright morning light from a window on the left, shot on a 35mm lens at chest height. Keep her face, her hair, her glasses and her earring identical to image 1.",
  "inputs": {
    "referenceImages": [
      "https://assets.runware.ai/guides/ideogram-4-6/character-consistency/reference-rosa.jpg"
    ]
  },
  "width": 3072,
  "height": 2048
}
Response
[
  {
    "taskType": "imageInference",
    "taskUUID": "9a4f1c68-2e57-4b3d-8d90-6c1e7f3a2b45",
    "imageUUID": "3e7b2d9f-4c18-4a65-b1e0-8d5f2a6c9e31",
    "imageURL": "https://im.runware.ai/image/os/a14d18/ws/2/ii/3e7b2d9f-4c18-4a65-b1e0-8d5f2a6c9e31.jpg"
  }
]

The prompt has two jobs. Most of it describes what is new: where she is, what she is doing and how the shot is lit. One closing sentence names what to hold by category (her face, her hair, her glasses and her earring) and leaves the details of each to the portrait. width and height set the new frame, so a square portrait can become a wide ad or a story.

One character across a campaign

Each image below used the same portrait as its only reference, with the same closing sentence:

She stays the same person in every frame, across four placements and four kinds of light. Name the glasses and the earring because they are small, and naming a small detail points the model at it.

The outfit is part of the reference until the prompt says otherwise. The workshop prompt says nothing about clothing, so she keeps the denim shirt from the portrait. The garden prompt puts an apron over it, and the testimonial prompt replaces it with a cardigan "instead of her shirt". Say whether the new piece goes over the old one or replaces it, because "wearing a cardigan" alone leaves the model to decide whether the shirt stays underneath.

Two characters in one scene

A second portrait goes in as image 2. Name each person by their image number, and give each one a keep clause of their own.

Built from 2 references
  • Image 1
  • Image 2

The action gives each of them a separate job (she holds the panel, he drives the screw), so the model has a reason to keep two distinct people instead of blending them. "Each in the clothes from their own image" settles the outfits in one phrase.

Shoot the references the same way, head and shoulders in even light on a plain backdrop, so neither portrait brings a scene of its own into the result.

An illustrated mascot

The same mechanism holds a drawn character, where the reference also carries the style, from the outline weight to the palette. Start from a character sheet: the mascot alone, facing forward, on white.

Each prompt ends with "in the exact flat style and palette of image 1". For a drawn character the style is part of the identity, and naming it stops the model from redrawing the owl with shading or a richer palette than the sheet. The lettering on each format comes from the same request, so a mascot and its copy ship as one image.

Tips

  1. Start from a clean portrait. Head and shoulders, a plain backdrop and even light give the model the most to hold and the least scene to carry over.

  2. Spend the prompt on the new scene. The portrait carries the person, so the prompt describes the place, the action and the shot.

  3. Name what to hold by category. "Her face, her hair, her glasses and her earring" points at the features and leaves their details to the image.

  4. Name the small features. List glasses, earrings and a single head feather in the keep clause, so the model looks for them.

  5. Say what happens to the outfit. "Over her shirt" and "instead of her shirt" are different instructions.

  6. Give two characters two jobs. Separate actions and separate keep clauses keep them from blending.

  7. Ask for the sheet's style on a mascot. "In the exact flat style and palette of image 1" keeps the outline and the colors on model.